Sock Photography: The Exceptions Amazon Wrote for Socks

Sock photography is product photography for an item you shoot as a pair and sell as a pack. Amazon’s 2019 Clothing and Accessories style guide names socks as an exception to three separate main image prohibitions: mannequins, packaging and swing tags, and the single product rule. Many sock listings use none of them.

Every guide that ranks for this topic opens the same way. Diffuse the light, stop down to f/8, sweep a white background, keep props minimal. That advice is fine and it is not what loses listings. Brands follow all of it, shoot one beautiful sock on white, and put a single pair in the slot that has to sell a six-pack.

This guide takes the other route. What Amazon’s own documents say about socks, where two of those documents flatly contradict each other, why a 200-needle dress sock is a resolution problem before it is a lighting one, and which frames survive a generative pass.

Key Takeaways

• Socks are Clothing in Amazon’s taxonomy, not Accessories. The 2019 style guide lists “Socks and Hosiery” under Women’s and Girls’ Clothing and under Men’s and Boys’ Clothing, which changes which rules apply.
• Three main image prohibitions carve socks out by name. Mannequins and digitally modified images, packaging and swing tags, and the single-product rule all read “except for stockings or socks” or “exception for multi-pack underwear or socks” (Amazon Clothing and Accessories Style Guide, 13 December 2019).
• A 2021 Amazon guide reverses two of them and never mentions socks. It lists visible and ghost mannequins and packaging as not allowed, and puts multipacks off-model. Amazon’s current clothing help page still shows socks as an exception for mannequins and packaging in its image examples.
• Amazon’s own worked example for package quantity is a sock six-pack. One bag of six pairs is IPQ 1. Five bags is IPQ 5, for 30 pairs. Get it wrong and the listing misprices.
• A dress sock knits finer than most sweaters. A 200-needle machine on a 3.75 inch cylinder is roughly 17 needles per inch, on a garment that occupies a fraction of a thumbnail.
• A foot model is a separate line on a studio quote. The on-foot frame is also the one most likely to come back wrong from a generative pass, so keep one real capture.
• The listing crop and the ad crop are different files. Meta’s ads guide recommends 4:5 at 1440 x 1800 for Facebook and Instagram feed image ads (facebook.com/business/ads-guide, September 2026). Amazon’s clothing page asks for a main image close to 3:4, and Shopify says square images display best at 2048 x 2048. Compose tall, then crop down.
• Generated imagery has a measured upside and a measured limit. Generated backgrounds lifted click-through roughly 15% over baseline on a mostly-apparel catalogue, in a range of 4 to 40% (RecSys ‘24, arXiv:2408.12392). In Baymard’s usability testing, users judged cheap-looking images harshly, including products on mannequins, and Baymard advises virtual models only as a last resort (baymard.com, December 2020).

What is sock photography?

Sock photography is the image set that sells hosiery online: a pack laydown showing the full quantity, a single pair flat for the colourway, a cuff and heel detail crop, an on-foot or on-leg frame for scale and fit, and a packaging shot where the band or box is part of the purchase. The set answers three questions a shopper cannot resolve from a thumbnail. How many pairs arrive, how thick is the knit, and where does the sock sit on the leg.

That last question is why height belongs in the frame rather than only in the copy. No-show, ankle, quarter, crew and over-the-calf are the size grid for this category, and a shopper buying a crew sock is buying a length. A photograph settles it faster than a bullet point.

Socks also carry an unusual amount of structure for their size. The cuff has its own rib, the leg often carries the pattern, the heel and toe are reinforced in a different yarn, and athletic pairs add an arch band. Each of those is a selling point, and none of them is visible when a sock lies flat and empty.

The gap: you photograph a pair, you sell a pack

Nearly every hard problem in this category traces to one mismatch. The unit in front of the camera is a pair. The unit in the cart is a three-pack, a five-pack or a gift box.

That gap does not exist for most apparel. A shopper buying a jacket sees the jacket. A shopper buying socks sees one pair photographed beautifully, reads in the title that six are coming, and has to trust that the other five look like the one in the picture. Trust costs conversions, and it costs returns when the pack turns out to be five plain pairs and one patterned.

The second half of the gap is shape. A sock has no form until something fills it. Lay a crew sock flat and the cuff rib compresses, the heel gore folds into the profile, and the arch band disappears. The garment loses three of the features you paid the mill to knit in.

Amazon has answers to both halves, written down, and they are more permissive than most sellers assume.

What Amazon’s own guides say about sock images

Two older Amazon guides cover this category and they do not agree. Amazon’s current help pages in Seller Central add a third answer. Read them directly rather than through a summary. The 2021 guide says on its pages that it is “for informational purposes only,” so treat the older guides as background and confirm your own subcategory in Seller Central before a launch.

The 2019 style guide, where socks are named

The Clothing and Accessories Style Guide dated 13 December 2019 sets the baseline. Main images need a pure white background at RGB 255, 255, 255. The product must fill 85% or more of the image. Images should be at least 1500 pixels on the longest side at 72dpi, which is what switches on zoom. Amazon accepts JPEG and TIFF, and prefers JPEG.

Then the classification that decides everything else. The guide lists “Socks and Hosiery” inside Women’s and Girls’ Clothing, and “Socks” inside Men’s and Boys’ Clothing. Socks are Clothing, not Accessories. The rule for that group reads: “Women’s & Men’s Clothing must be photographed on model or flat.” Accessories get a stricter line, “must be photographed flat including scarves, belts, hats, ties & cold weather accessories,” and socks are not in that list.

Three prohibitions then name socks as exceptions, verbatim:

• “Main image must be a single product, not stacks of images containing multiple products. Exception for multi-pack underwear or socks.”
• “Product images photographed on mannequin or digitally modified images, except for stockings or socks.”
• “Images that show packaging or brand or swing tags, except for stockings or socks.”

Read together, that is a permission set no other garment gets. A sock listing may show the whole pack in the main image, may use a visible foot or leg form, and may show the band or box the pairs arrive in. Those three are precisely the images a sock shopper wants, and they are the three most sock listings leave out.

The same guide uses socks as its worked example for Item Package Quantity. One bag of “Hanes Men’s 6 pack cushion crew socks” is IPQ 1, because the customer receives one package containing six pairs. Five of those bags is IPQ 5, for 30 pairs total. That field drives pricing and quantity display, and this category is where it goes wrong most often.

The Spring 2021 guide, where socks disappear

The Seller Imaging Guide for Clothing and Accessories, Spring 2021 edition, reverses two of the three exceptions and never mentions socks or hosiery anywhere in the document.

It raises the resolution floor to at least 1,600 pixels in height with a 10,000 pixel ceiling on the longest side. It keeps the 85% fill and the pure white background. Then it says adult apparel should be imaged on-model, and multipacks and sets off-model. Its Main Image DON’Ts list includes visible mannequins, ghost mannequins, showing packaging or swing tags, and multiple product views in a single image. It adds: “Note: mannequins are not allowed in product images.” The page closes with the enforcement line, that failure to meet the requirements “may result in the removal of your ASIN from search.”

The one exception it preserves is the multipack carve-out: main images “must be a single product, not stacks of images containing multiple products (except multipacks/sets).”

Amazon’s current help pages, September 2026

Amazon’s current Seller Central pages sit between the two guides. The clothing image guidelines page asks for main images on a model, with exceptions, and lists a visible mannequin and visible packaging or tags among prohibited main images. Its image examples then add “except for stockings or socks” to the rules on visible mannequins and on packaging and brand tags. A main image may show several products only when they are sold together as a multipack named in the title (sellercentral.amazon.com, September 2026).

The general product image guide is stricter on one point. It says to show no part of a mannequin, and to show clothing accessories or multipacks flat, without a model. It sets a 500 pixel minimum on the longest side, and zoom turns on at 1,000 pixels (sellercentral.amazon.com, September 2026).

The practical read: the multipack laydown is the one shot all three sources agree on, so build the listing around it. The foot form and the packaging shot are allowed by the 2019 guide and the current clothing examples, but not by the 2021 guide or the product image guide’s mannequin rule. That makes them variant-slot material rather than a main image you bet an ASIN on. Our clothing photography shot list covers the on-model rules that govern the rest of an apparel range.

One more warning worth carrying. Some third-party “Amazon image requirements” pages say ghost mannequins are fine for adult clothing main images. Amazon’s Spring 2021 guide lists ghost mannequins under DON’Ts, and the current product image guide says to show no part of a mannequin. Check any claim about this against the source document, including the ones on this page.

How to lay a pack so the count reads at thumbnail size

The March 2022 Category Style Guide publishes the composition for a multipack laydown, and it reads like a brief from an art director.

The products should be shot flat using pins and liner or stuffing of your choosing, using seams as a guideline, symmetrical left to right, steamed and free of loose threads. They should be laid on top of one another cascading diagonally from left to right, darkest product top left grading to lightest for lighting purposes, or with the print featured at the front.

Three instructions in there each solve a real sock problem.

“Pins and liner or stuffing of your choosing.” This is permission to give the sock shape without a mannequin. A cut card liner inside the leg holds the cuff open and keeps the rib legible, which is the off-figure answer to a garment that collapses when empty.

“Cascading diagonally from left to right.” A diagonal cascade shows the count without merging the pairs. Six pairs squared into a grid read as texture at 300 pixels wide. Six pairs fanned on a diagonal read as six.

“The darkest product should be top left to lightest.” A tonal ramp stops the darkest pair swallowing the shadow side of the frame and gives the eye an order to follow. Lead with the printed pair instead when the print is the reason someone buys the pack.

Match the number of pairs in the frame to the number in the title, and match both to the IPQ field. A listing titled six-pack showing four pairs is a returns problem created in the studio. The general craft of the laydown, lighting and squaring, sits in our flat lay photography guide.

The needle count that sets your detail crop

Sock machines are specified by needle count on the cylinder, and that number predicts how a knit will photograph. One maker’s specification for its RB-6FTP machine, with a 3.75 inch cylinder, maps needle count to sock type: 96N and 108N for baby socks, 120N for children’s socks, 132N for teenagers, 144N for ladies’ or men’s socks, 156N and 168N for men’s socks, and 200N for what the maker calls quality men’s socks (rbsockmachine.com, September 2026).

Run the arithmetic at the top of that range. 200 needles around a 3.75 inch cylinder works out at roughly 17 needles per inch. On the knitwear gauge scale, 12GG to 18GG is the fine band, the one that triggers moire and mushes under compression. A dress sock sits inside it.

The difference is scale. A fine-gauge sweater fills the frame at 2048 pixels, so each stitch still gets several pixels. A sock occupies a fraction of a small thumbnail, so the same knit resolves to nothing. The pattern you paid for averages into flat colour exactly where the shopper first sees it.

The fix is the one that applies to any fine-gauge knit. Put more pixels on each stitch. Move closer, or output larger. Nothing about the lighting changes the arithmetic. Where an existing frame is too small, an upscale app can enlarge it before the storefront resamples it down. It cannot add stitch detail the camera never caught.

Shopify sets the other end of this. Product images can run to 5000 x 5000 pixels or 25 megapixels, under 20 MB, and Shopify says 2048 x 2048 usually displays best for square product images. A product can hold up to 250 images, videos and 3D models combined (help.shopify.com, September 2026). A sock brand with 40 colourways at six frames each uses 240 of those 250 slots, which leaves no room for a video per colourway. That is the practical argument for splitting colourways across products rather than variants.

The on-foot shot, and why it carries the most risk

An on-foot frame answers what a laydown cannot. Where the cuff lands, how the heel sits, whether the pattern distorts when the sock stretches. Amazon’s 2019 guide permits it for socks either on a model or on a form. It is also the frame most likely to come back wrong from a generative pass.

Feet are five similar elongated shapes in close contact, frequently overlapping. That is the same condition that makes hands hard for image models. The failure modes look like the ones that show up in hands and faces: a toe count that does not survive a zoom, a heel that bends the wrong way, an ankle that reads as putty where the cuff meets skin.

Three rules keep it manageable.

Frame out what you do not need. A crew sock listing needs the ankle and calf, not the toes. Crop at the arch and the hardest region leaves the frame entirely.

Keep feet planted and separated. Standing, seated on a bench, one foot forward on a step. Crossed ankles and tucked-under feet stack overlapping shapes and multiply the error rate.

Check every generated frame at full size before it ships. Toe count, cuff line, and where the sock meets skin. Seconds per frame, and the only reliable gate.

There is an honest counterweight to generating this frame at all. In Baymard Institute’s usability testing on apparel and accessories, users judged cheap-looking images harshly, including products shown on mannequins. Baymard advises mannequins or virtually rendered models only as a last resort, because shoppers use model images to judge fit, tightness and how fabric drapes (baymard.com, December 2020). That research predates current models by more than five years and is qualitative rather than measured, so treat it as a risk signal rather than a verdict. It is still the reason to keep at least one real on-foot capture in a set. For the general case, our guide to on-model photography with AI covers where the verification effort pays.

The sock shot list

Six frames cover a sock listing. Amazon’s product image guide recommends a main image plus at least six more images and one video (sellercentral.amazon.com, September 2026). Its clothing page allows up to eight alternate images per SKU. The six frames below are the core, and extra colourways or a video fill the rest.

Order matters more for socks than for most categories, because the assortment question outranks the aesthetic one. A shopper who cannot tell what is in the pack does not get as far as admiring the knit.

The same pair, three crops

The listing frame and the ad frame are not the same file. Marketplace and store images sit at 3:4 or square. Paid social wants vertical. A sock brand that ships one square export to every placement is padding dead space into the placement that gives it the most screen.

Meta’s own ads guide recommends 4:5 at 1440 x 1800 pixels for both Facebook Feed and Instagram Feed image ads (facebook.com/business/ads-guide, September 2026). Amazon’s clothing page asks for a main image with an aspect ratio close to 3:4, and Shopify says 2048 x 2048 usually displays best for square product images.

Socks are the category where this ordering matters. A sock is a tall narrow object, and a pair on a leg is taller still. The 4:5 crop fits the garment. A square crop is the one that throws frame away at the top and bottom.

So compose tall and crop down for the 3:4 and square slots. Going the other way pads a square export into 4:5 and puts a small sock in the middle of empty background, at the exact placement where vertical height was the reason to buy the impression. Set the target size before the export rather than after, or run the finished frame through an image resizer with the placement preset picked first.

Keeping a colourway drop consistent

Sock brands do not ship one product. They ship twelve colourways of the same crew sock, and the range is judged as a grid rather than as twelve separate pages.

That is where drift shows up. Shoot or generate each colourway on its own and the cuff angle moves, the light shifts a few degrees, the crop tightens by five percent. Every frame passes on its own. Lined up in a collection page they read as twelve different products from three different brands.

The fix is composition first, colour second. Lock one source pair, one camera position and one crop, then change only the yarn colour across the range. Everything that is not the colourway should be identical by construction rather than by eye.

Two habits make it hold:

Save the brand inputs once. You set the palette and the rules once in a brand profile. The workflow reads it on every run, so colourway nine starts from the same place as colourway one instead of from a rewritten prompt.

Review the range as a contact sheet, not as files. Put all twelve at thumbnail size in one grid before anything ships. Drift is invisible at full size and obvious at 300 pixels wide, which is also the size the shopper sees.

How a sock shoot is priced

We could not find a survey of sock photography prices, so this section describes how studios meter the work rather than quoting a number.

Now put the per-image model against the product. A sock brand carrying 30 colourways at six frames each needs 180 images, and a per-image rate card bills every one of them. The bill grows with the frame count, whatever a pack of socks sells for.

A handbag brand pays per frame too, on a much higher order value. For a low-price pack of socks, the photo bill takes a larger share of what each sale earns. Our full breakdown of what a product photoshoot costs covers the general case.

Where AI holds up on socks, and where it breaks

There is real measured evidence on both sides, and it is narrower than most claims about it.

On the upside, a controlled study across catalogues of a few thousand to several tens of thousands of items, most of them apparel, found products with generated backgrounds gained roughly 15% in click-through over baseline, with the range across merchants and placements running 4 to 40%, all significant at p below 0.05 (Czapp et al., RecSys ‘24, arXiv:2408.12392). Scope that honestly: it measures generated backgrounds on recommendation and ad creative, not product detail page heroes, and it says nothing about how many images a listing needs.

On the post-purchase question, randomised field experiments with millions of users at a cross-border retail platform found sales effects from generative AI running from nothing up to 16.3%, driven by conversion rather than basket size. The authors report that “product return rates and customer ratings do not deteriorate” (Fang et al., arXiv:2510.12049, revised June 2026). The study covers generative AI across seven workflows, including advertising and seller services, and not imagery alone, so treat it as indirect evidence on returns.

Against that, discount every “professional photos increase conversion 94%” and “each extra image adds 5 to 8%” figure in circulation. We could not trace any of them to a published study or a stated method.

The category split for socks specifically:

Holds up well. Colourway variants from one photographed pair, which is where the volume cost sits. Pack laydowns rearranged into different assortments. Background and surface changes for social and paid placements. Angle variations on a single pair.

Holds up with checking. Jacquard and argyle repeats. A model reproduces the look of a diamond pattern rather than the logic of one, so the repeat can drift across the leg or break at the heel. Compare against the real garment before it ships.

Breaks. Anatomy in on-foot frames. Legible text on a cuff or a woven band, which is a different problem to solve than pattern. Anything where the exact pair count has to be right, because a generated pack is a composition a model will happily edit.

On model choice, one provider documents the property this category needs. Black Forest Labs describes FLUX.2 [flex] as “Specialized for typography and text rendering, and for preserving small details” (docs.bfl.ai, September 2026). That is a vendor statement, so test it on your own sock before you trust any model with knit texture. The model list shows the image models DesignerBox runs, and running one brief through two of them is the quickest comparison.

The rule that holds across all three tiers: keep one real photograph of the actual sock in the source set. Everything derived from it inherits the true colour, the true rib and the true yarn. Everything generated from a text description inherits a guess.

Six mistakes that cost a sock listing

Each of these is cheap to fix and none of them is a lighting problem.

One pair in a multipack main image. The listing sells six and the frame shows one, in a slot where every Amazon source allows a multipack. Match the count in the frame to the count in the title and to the Item Package Quantity field.

Checking the detail crop at full size. A jacquard that reads beautifully at 2048 pixels averages into flat colour in the grid. Judge every pattern frame at thumbnail size, because that is where the shopper meets it.

Shipping a generated on-foot frame unchecked. Toe count, heel direction and the line where the cuff meets skin. Seconds per frame, and the failure is the kind a customer screenshots.

Reusing the square listing crop as the ad creative. It pads into 4:5 and puts a small sock in the middle of empty background. Compose tall, then crop down for the marketplace.

Letting colourways drift. Twelve pairs shot twelve separate times read as three brands in one collection grid. Lock the composition and change only the colour.

Photographing a hand-linked toe so the closure reads as a bulky seam. Brands charge a premium for that construction and then shoot it from the angle that contradicts the claim. Turn the toe, light it flat, and let the detail crop prove what the copy says. Same logic for reinforced heels: if durability is the argument, one frame has to carry it.

A sock set in DesignerBox

The workflow starts with one photographed pair and ends with the full listing set.

1. Shoot one pair properly, large. Flat, lit from the side so the rib casts, plus a macro at the pattern. This is the source everything else derives from, so it is the frame worth paying a studio for.
2. Build the pack laydown. A flat lay template composes the multipack from your pairs. Set the cascade and the tonal ramp there. Match the count in the frame to the count in the title and the IPQ field.
3. Run the colourway range. Run the workflow once for each colourway, from the same source pair. Check each colourway against the real yarn before it ships.
4. Fill the variant slots. A multi-angle template makes the cuff, heel and profile views from the same source.
5. Verify the frames that carry rules. The main image against 85% fill and the white background. Every on-foot frame against anatomy. Every jacquard repeat against the real sock.

You see the cost of a run before you start it. The free plan includes 112 credits a month and takes no card. Basic is $15 a month billed monthly for 500 credits, Pro is $35 a month billed monthly for 1,000, and Premium is $75 a month billed monthly for 2,500. The commercial licence starts at Pro. Current terms are on the pricing page.

The point of steps 1 to 5 is that you set them up once. Save the sequence as a workflow. The saved workflow runs the same way on the next colourway, the next pack size and next season’s range, so colourway one and colourway thirty come from the same steps. Batch, which will run one workflow over a whole sheet of products, is coming. To build the first one, start from the templates. The cost is shown before the run.

FAQ

How do you photograph socks for Amazon?

On a pure white background at RGB 255, 255, 255, with the product filling 85% of the image. Amazon’s 2019 style guide classifies socks as Clothing, permits them “on model or flat,” and names socks as an exception to the mannequin, packaging and single-product prohibitions. The Spring 2021 imaging guide reverses the first two and does not mention socks. Amazon’s current clothing page keeps sock exceptions in its image examples, and its product image guide says to show multipacks flat, without a model. Build the listing on the pack laydown, which all of them allow.

Can you use a mannequin or foot form for sock photos on Amazon?

Amazon’s 2019 Clothing and Accessories style guide lists as prohibited images “photographed on mannequin or digitally modified images, except for stockings or socks,” which is an explicit carve-out. The Spring 2021 imaging guide states that mannequins are not allowed in product images and lists both visible and ghost mannequins under Main Image DON’Ts, without mentioning socks. Amazon’s current clothing page shows visible mannequins as not allowed except for stockings or socks, but its product image guide says to show no part of a mannequin. Treat the foot form as variant-slot material and confirm your own subcategory in Seller Central.

Should sock photos be flat lay or on a foot?

Both, in different slots. The pack laydown is the shot every Amazon source allows, so it takes the main image. On-foot frames answer height on the leg and fit under stretch, and belong in the variant slots. A direct-to-consumer store with no marketplace rule can lead with either, though the pack shot still resolves the assortment question faster.

How do you make a sock hold its shape without a mannequin?

Use a cut card liner or light stuffing inside the leg, which Amazon’s multipack crop guide explicitly permits: shoot flat “using pins and liner or stuffing of your choosing,” symmetrical left to right. It holds the cuff open so the rib stays legible and keeps the heel gore from folding into the profile, without putting a mannequin in the frame.

How many photos does a sock listing need?

Six covers it: pack laydown, single pair flat, cuff and rib crop, heel and toe detail, on-foot or on-leg, and a packaging shot. Amazon’s product image guide recommends a main image plus at least six more images and one video, so add a video or extra colourways on top of the six.

Why do my sock patterns disappear in the product grid?

Resolution, not lighting. A 200-needle dress sock knits at roughly 17 needles per inch, and a sock occupies a fraction of a thumbnail, so each stitch resolves to less than a pixel and the pattern averages into flat colour. Shoot the detail crop closer and output larger, then check the garment at thumbnail size rather than at full size.

Do sock multipacks need to show every pair?

Amazon’s guides allow a multipack main image to show multiple products. The current clothing page asks that the multipack be named in the title, and the count in the frame should match the count in the title and the Item Package Quantity field. Amazon uses a sock six-pack as its own worked example: one bag of six pairs is IPQ 1, five bags is IPQ 5. Showing one pair on a six-pack listing wastes a slot you are permitted to use.

What size should sock photos be for Instagram and Facebook ads?

Meta’s ads guide recommends 4:5 at 1440 x 1800 pixels for image ads in both the Facebook and Instagram feed. That is taller than the 3:4 main image Amazon’s clothing page asks for and the square Shopify says displays best, so export twice rather than padding one file into both. Socks favour the vertical crop, since the garment is tall and taller again on a leg, so compose for 4:5 and crop down for the marketplace.

How do you keep twelve sock colourways looking like one range?

Lock the composition and change only the yarn colour. One source pair, one camera position, one crop, with every colourway derived from that frame rather than shot on its own. Then review all twelve together at thumbnail size before anything ships, because framing drift is invisible at full size and obvious in a collection grid.

Can AI generate sock photos that keep the real knit?

For colourway variants from a photographed pair, generally yes, because the knit face reads as a consistent surface and the source frame carries it. For jacquard and argyle, verify each frame, since a model imitates the pattern rather than following its construction. For anything with feet in it, check the anatomy at full size before it ships.

Sources

• Amazon Seller Central, Product image guide: mannequins, multipacks, image size, zoom and recommended image count, accessed September 2026
• Amazon Seller Central, Image guidelines for clothing: main image rules, the stockings and socks examples, aspect ratio and alternate images, accessed September 2026
• Amazon, Category Style Guide: Clothing and Accessories, dated 13 December 2019: garment taxonomy, main image requirements, the stockings and socks exceptions and the IPQ example, accessed September 2026
• Amazon, Seller Imaging Guide: Clothing and Accessories, Spring 2021: main image DO’s and DON’Ts, on-model rule and resolution floor, marked informational only, accessed September 2026
• Amazon, Selling Partner Style Guide: Clothing and Accessories, March 2022: multipack crop guide, accessed September 2026
• Shopify Help Center, Adding product media and Product media types, accessed September 2026
• Meta Ads Guide, image ad recommendations for Facebook Feed and Instagram Feed, accessed September 2026
• Rainbowe RB-6FTP specification: needle counts, sock types and cylinder size, rbsockmachine.com, September 2026
• How studios meter photos and models: soona.co/pricing and squareshot.com/pricing, September 2026
• Czapp, Jani, Domian and Hidasi, Dynamic Product Image Generation and Recommendation at Scale for Personalized E-commerce, RecSys ‘24 Industry Track
• Fang, Yuan, Zhang, Donati and Sarvary, Generative AI and Sales Productivity: Field Experiments in Online Retail, revised June 2026
• Baymard Institute, Provide Images of Accessory, Apparel, and Cosmetic Products on a Human Model, published 1 December 2020, accessed September 2026
• Black Forest Labs, FLUX.2 [flex] API documentation, accessed September 2026
• DesignerBox pricing page (designerbox.ai/pricing), September 2026

Marketplace image rules verified against Amazon’s current Seller Central pages and its December 2019, Spring 2021 and March 2022 clothing guides as of September 2026. The older guides contradict each other and the current pages on mannequins and packaging, so confirm your own subcategory in Seller Central before a launch. Individual results vary.